Professional Documents
Culture Documents
Chemical Bond
Chemical Bond
Chemical Bond
D.N. Talwar
9.1 Introduction
In the recent years, dilute ternary and quaternary III–As–N alloys have
received a great deal of attention [1–9] both from a fundamental point of view
as well as for applications in technology (e.g., photodetectors, modulators,
amplifiers, and long wavelength vertical cavity surface-emitting lasers (LW-
VCSELs), etc.). Dilute nitrides can be derived from the conventional III–V
semiconductors (viz., GaAs, GaInAs or GaAsP) by the insertion of N into the
group V sublattice – causing profound effects on the electronic properties of
the resulting alloys [10–14]. Contrary to the general trends in the conventional
alloy semiconductors where a smaller lattice constant generally increases the
band gap, the smaller covalent radius and larger electronegativity of N cause
a very strong bowing parameter in III–V–N compounds. Consequently, the
addition of N to GaAs or GaInAs decreases the band gap (Eg ) dramatically
(see Fig. 9.1). This strong dependence of Eg on the N content in III–As–N
has provided opportunities to engineer material properties suitable for the
224 D.N. Talwar
3.2 GaN
GaP
AIAs
2.0
Bandgap (eV)
1.5 GaAs
x
-x N
InP
1.3 µm
s1
1.0
GaA
1.55 µm
0.5
Nx InAs
s 1-x
InA
0.0
4.4 5.4 5.5 5.6 5.7 5.8 5.9 6.0 6.1 6.2
Lattice constant (Å)
Fig. 9.1. Band gaps vs. lattice constants of binary III–V semiconductors (squares).
The continuous and dashed lines indicate direct and indirect band gaps of ternary
compounds, respectively. The areas between lines represent quaternary compounds.
The Ga- and In-rich GaInNAs materials can be grown lattice matched to GaAs and
InP [3], respectively, to reach the fibre-optical communication wavelengths of 1.3
and 1.55 µm
host lattices and modify their vibrational characteristics as well. If the impu-
rity atoms are sufficiently light compared to the host atoms, then some of the
modified modes may occur at frequencies above the maximum normal mode
frequency (ωmax ) of the perfect lattice. These new high frequency modes can-
not propagate through the lattice and are highly localized spatially around the
impurity site. Because of the spatial localization of the vibrational energies
of such impurity modes, they are called “localized vibrational modes”, here-
after designated by LVMs. If the modified modes occur between the phonon
branches of the perfect lattices where the phonon density of states is zero,
then they are also spatially localized and are called “gap modes”. Of special
interest here are the high frequency LVMs since, as will be shown, they occur
in the material systems being studied. In general, these modes have a non-
zero dipole moment associated with them, and are, therefore, infrared active.
Besides IR absorption, Raman scattering spectroscopy [23–29] has been con-
sidered equally powerful technique with strong ability to sense not only the
electronic properties but also to identify the nature of impurities and their
bonding mechanisms in semiconductors.
BAs , and SiAs in GaAs have also been reported by using the high-resolution
Fourier transform infrared (FTIR) spectroscopy [27].
All these studies with improved accuracies in the experimental results have
provided strong spectroscopic “fingerprints” for the site selectivity of impurity
atoms and contributed to the fundamental understanding of lattice dynamics
as well as impurity–host interactions (i.e. chemical bonding) in the cases of
both isolated and complex defect centres in semiconductors. Experimental
results have also been tested and interpreted rigorously in the framework of
realistic lattice dynamical schemes by using sophisticated Green’s function
and other methods [37].
Until recently, the information about the LVMs of nitrogen in III–V com-
pounds was rather sparse. In GaAs or InAs the addition of a small amount
of nitrogen occupying the anion site (NAs ) acts as an isoelectronic impu-
rity [10–14]. For a light N atom on the heavier As site, one expects to
observe an LVM of NAs by using either IR absorption or Raman scattering
spectroscopy.
228 D.N. Talwar
Infrared Absorption
In GaAs, the first spectroscopic evidence of nitrogen local mode was reported
by Kachare et al. [38] who observed a broad absorption band near 480 cm−1
after implanting high energy 14 N+ ions into semi-insulating GaAs. More
recently, it has been shown that the NAs local mode can be observed in epi-
taxial layers of GaAs1−x Nx with thickness as small as of 10 nm, in the alloying
range, for the N fraction of x < 0.03. IR measurements performed in nitrogen-
doped GaAs prepared by vapour phase epitaxy as well as in GaAs1−x Nx
layers grown by liquid source MBE have revealed NAs -related LVM near
∼470 cm−1 [31]. This assignment has now been confirmed, with the implan-
tation of 15 N in GaAs and observing shift of the local mode towards lower
∼458 cm−1 frequency. It is worth mentioning that in an earlier study [30] the
NAs local mode near ∼471 cm−1 was detected in a nominally undoped bulk
GaAs crystal where nitrogen was a contaminant (∼1015 cm−3 ) coming from
either the N2 gas or the pyrolithic boron nitride (pBN) crucible used for the
crystal growth.
In Fig. 9.2, we have displayed the FTIR absorption spectrum (77 K) [30]
of a GaAs0.983 N0.017 film grown by MOCVD on a semi-insulating GaAs sub-
strate. Clearly, the absorption spectrum after subtracting the reference spectra
of semi-insulating GaAs has revealed a relatively broad (∼13 cm−1 ) band of
the NAs local mode near ∼472.5 cm−1 . At the same time the N-related mode
exhibited some very peculiar behaviour. If comparison is made with the LVMs
of closest mass 11 BAs and 12 CAs in GaAs (a) the frequency of the NAs local
mode is relatively lower (Table 9.1) and (b) since the nearest-neighbour atoms
1.0 2TO
GaAs1-xNx
x = 0.017
0.8 d = 60 nm
77 K
TO+LA
Absorbance
0.6
(a)
0.4
472.5
0.2
13.2
(b) ⫻ 10
0.0
400 450 500 550
Wavenumber (cm−1)
of NAs are gallium atoms, the local mode does not reveal the fine structure
splitting [30] due to Ga-isotopic (69 Ga and 71 Ga) masses [39, 40]. Again, the
observed broad bandwidth of NAs -mode is likely to be caused by the reduced
lifetime of the vibrational state and it is assumed that elastic scattering [41]
of the lattice phonons might be a dominant process.
In InAs, the FTIR measurements performed on 14 N- and 15 N-implanted
bulk materials have also revealed LVM frequencies near ∼443 and 429 cm−1 ,
respectively [30].
Raman Scattering
T = 77 K
T = 300 K
Fig. 9.3. Raman spectra of GaAs0.99 N0.01 layer on GaAs recorded at 77 and
300 K [35]
230 D.N. Talwar
Ga(As,N) Ga(As,N)
sample A sample A LO2
hνL = 1.92 eV
LO2
TO2 TO2
1.83 eV
X Y B = 14 T
Intensity (arb. units)
(b) B=0 T
1.92 eV T = 75 K
source MBE at 500◦ C, where rf-plasma source is used for the nitrogen supply.
The spectra, excited at three different incident photon energies (1.83, 1.92,
and 2.18 eV), are displayed in Fig. 9.4a, where as in Fig. 9.4b the effects of
the external magnetic fields and temperatures at a fixed excitation energy of
1.92 eV are revealed.
It is worth mentioning that the defect-induced Raman scattering efficiency
is stronger only for the excitation at photon energies in the range between 1.8
and 2.2 eV. From Fig. 9.4a, one can note the expected GaN-like LO2 phonon
line near ∼472 cm−1 superimposed on the background of the second-order
phonon scattering. A similar resonance behaviour of the LO2 phonon signal
in the energy range between 1.9 and 2.0 eV has previously been attributed
to the enhancement of the Raman efficiency for photon energies approaching
the localized E+ transition in GaNAs [35]. Based on the selection rules, the
peak near 463 cm−1 can be assigned to either the scattering by GaN-like TO2
phonon or two GaAs-like 2LA1 phonons. The Raman features at frequencies
higher than ∼500 cm−1 are related to the second-order scattering by GaAs-
like TO1 and LO1 phonons. Two additional peaks, observed in Fig. 9.4a near
∼409 cm−1 (line X) and 427 cm−1 (line Y ), are attributed to the N-related
vibrational modes (possibly N dimers on Ga and As sites, respectively) as the
external magnetic fields up to 14 T (see Fig. 9.4b) do not cause any split or
shift to these lines. Similar to GaAsN, Wagner et al. [42], while analysing the
bonding of nitrogen by Raman spectroscopy in InAsN, not only observed (see
9 Assessing the Preferential Chemical Bonding of Nitrogen 231
InAs1−yNy
442.7cm−1
hνL = 2.54 eV
T = 77 K InN-like
y = 0.005
y = 0.012
Fig. 9.5. Raman spectra in MBE-grown dilute InAs1−y Ny layers for the excitation
at 2.54 eV with 0 ≤ y ≤ 0.012 [42]
Fig. 9.5) the 14 NAs LVM ∼ 443 cm−1 but also detected two additional modes
possibly due to N dimers near 402 and 416 cm−1 , respectively.
It is well known that the electronic and optical properties of dilute quater-
nary Ga1−x Inx Ny As1−y alloys strongly depend upon the microscopic spatial
arrangement of its constituent elements. Unlike various other techniques (e.g.
PL, XRD) used for assessing the incorporation of nitrogen in semiconductors,
the local mode spectroscopy has been considered to be very well suited for
addressing the key issues of the local N bonding in GaAs (InAs)-based low-
In (Ga) content dilute Ga1−x Inx Ny As1−y alloys. This is simply because the
transverse optical phonon (TO) in cubic GaN, which is the symmetry for the
dilute Ga–N bonds in zinc-blende GaInAs-like host matrix, is nearly disper-
sion less. Consequently a small fluctuation in the Ga–N frequency can give
rise to phonon localization, i.e. to different signals corresponding to the many
zone-centre modes. In other words, the phonon frequencies in low-In (Ga) con-
tent dilute Ga1−x Inx Ny As1−y alloys will be strongly dependent on the local
bond arrangement related to the large differences in the bond strengths and
bond lengths between the Ga–N and (Ga,In)–As bonds.
In the recent years, several FTIR measurements on Ga1−x Inx Ny As1−y /GaAs
materials (0 ≤ x ≤ 1 and y < 0.02) grown by MOCVD and gas source
MBE techniques have been reported in the literature [30, 31, 43, 44]. Despite
a general consensus on the NAs local mode in GaAs and InAs, the results
232 D.N. Talwar
471.4
0.05
InGaAsN: 77 K
x = 0.018
0.04 5⫻10nm
88-grown
Absorbance
0.03 annealed
0.02
0.01
0.00
Fig. 9.6. Low-temperature FTIR absorption spectra in a Ga1−x Inx Ny As1−y MQW
structure before and after annealing [30]
472
458
0.01 0.01
447
460
478
492
Absorbance
Absorbance
450 490 440 480
1:1 1:1
1:2
1:2
14N 15N + 115In
1:3 + 115In 1:3
(a) (b)
Fig. 9.7. FTIR absorption spectra [44] of GaAs samples co-implanted with indium
and nitrogen isotopes (14 N and 15 N) after rapid thermal annealing at 800◦ C: (a)
co-implantation of 14 N and 115 In and (b) co-implantation of 15 N and 115 In
y = 0.07, x = 0.02
(a) GaAs0.99N0.01
9.3 Theoretical
There are two main theoretical methods available for the calculations of impu-
rity modes in semiconductors (1) the super-cell approach [45, 46] and (2) the
crystalline Green’s functions method [47,48]. These techniques differ primarily
in the treatments of the surrounding atoms of the defective region. In the ab
initio super-cell approach, one considers a large unit cell containing the defect
9 Assessing the Preferential Chemical Bonding of Nitrogen 235
In the ab initio super-cell approach [45, 46], the selected number of atoms in
the large unit cell are displaced to all three Cartesian directions. After each
displacement the electronic structure for the new configuration is optimized
and the resulting Hellmann–Feynman forces are calculated. This is done for
all the atoms which are a priori considered important for the description of
the impurity modes of the defect structure. The dynamical matrix is then
calculated by finite difference using the forces and displacements. The normal
modes and the corresponding vibrational frequencies are then obtained by
diagonalizing the dynamical matrix.
|I + Go P| = 0. (9.1)
Phonons
Defect Configurations
To comprehend the observed FTIR [30, 31, 44] and Raman [35, 42, 43] scat-
tering data on LVMs in Ga1−x Inx (Alx )Ny As1−y quaternary alloys (y ≤
0.03 and 0 ≤ x ≤ 1), one can select five defect configurations involv-
ing isolated elementary tetrahedrons related to the incorporation of nitro-
gen (see Fig. 9.9, viz., NInm (Alm )Ga4−m with m = 0, 1, 2, 3, and 4:
NGa4 (Td ), NIn1 (Al1 )Ga3 (C3v ), NIn2 (Al2 )Ga2 (C2v ), NIn3 (Al3 )Ga1 (C3v ), and
NIn4 (Al4 )(Td )). In GaIn(Al)NAs alloys, as the symmetry of the defect config-
urations is changing from Td → C3v → C2v → C3v → Td , one anticipates from
group theoretical arguments a total of nine optically active impurity modes.
Generally, the frequencies of impurity modes in mixed configurations fall
between the values of LVMs of the two end members, i.e. between the local
modes of NAs in GaAs (with x = 0 for In(Al)-free case) and In(Al)As (with
x = 1 for Ga-free case). Notable exceptions can occur, however, if the new
equilibrium positions of the defect centres in the alloy systems are associated
with significant variations in the force constants, as might be the case, in our
choice of Ga1−x Inx (Alx )Ny As1−y quaternary alloys.
9 Assessing the Preferential Chemical Bonding of Nitrogen 237
Fig. 9.9. Representation of the zinc-blende GaAs lattice with isolated NAs and
InGa defects (a), NAsInGa(1)Ga3 nearest-neighbour pair (b), NAsInGa(2)Ga2 next
nearest-neighbour pair (c), and NAsInGa(3)Ga1 complex centre (d)
Lattice Dynamics
The host crystal phonons are described by an RIM11 [51] in which the force
constants are written as the sum of a short-range bonding-type interactions
and a long-range interaction due to the Coulomb field. The short-range forces
are assumed to be zero after the second nearest neighbours, which when com-
bined with the zinc-blende symmetry conditions, restrict parameters to ten
[A, B (nearest neighbour); Ci , Di , Ei , and Fi ; i = 1, 2 (second nearest neigh-
bour)] force constants. The long-range interaction is derived by assuming that
238 D.N. Talwar
Table 9.2. Rigid-ion model force constants and Szigeti effective charge used for the
lattice dynamical (phonon) calculations of GaAs, AlAs, and InAs
atoms behave like rigid ions with an effective Szigeti charge Zeff . All the model
parameters are calculated by using a non-linear least-square fitting procedure
with constrained parameters and weighting to the available data on phonon
frequencies at critical points (Γ , X, L, and K) and elastic constants. The
parameters used (see Table 9.2) here for GaAs are those of Kunc et al. [51]
which model the phonon dispersions relatively well, providing an excellent
agreement with the neutron scattering data [36]. For AlAs and InAs, how-
ever, we have chosen the set of parameters from [54] and [48], respectively,
obtained by fitting to the existing IR and Raman scattering data.
As a representative case, we have displayed (see Fig. 9.10) the calculated
one-phonon density of states and the phonon dispersions of GaAs (using the
parameter set from Table 9.2) compared with the neutron scattering data
of Strauch and Dorner [36]. This serves to illustrate the level of agreement
obtained by our lattice dynamical model (RIM11). Moreover, we find a very
good concurrence in the phonon frequencies at high-symmetry points when
comparison is made with the values obtained by more sophisticated ab initio
methods [55]. It is worth mentioning, however, that a set of force constants in a
phenomenological scheme providing good agreement to the neutron scattering
IR and Raman data of phonon frequencies of a perfect crystal is not necessarily
a guarantee for the accuracy of the model.
The point that compels recognition to a lattice dynamical scheme demand
simultaneously correct eigenvalues and eigenvectors. Although the impetus for
examining the phenomenological models comes from ab initio [45, 46] meth-
ods, the later approaches have not yet replaced the former schemes completely.
The simple reason is that the ab initio methods deal only with the phonons at
a few high-symmetry points while the efficiency of phenomenological models
9 Assessing the Preferential Chemical Bonding of Nitrogen 239
300
250
Frequency (cm−1)
200 GaAs
150
100
50
0
Γ ∆ Χ Σ Γ L
DOS (arb. units)
Wavevector, q
Fig. 9.10. Rigid-ion model calculations for the phonon dispersions along high-
symmetry directions compared with the neutron scattering data [36] and one-phonon
density of states obtained with the parameter values of Table 9.2 for GaAs
comes to profit when different Brillouin zone averages are evaluated, viz., in
the calculations of Green’s functions used for studying the impurity vibra-
tions [37]. Our earlier investigation of LVMs for identifying the microscopic
lattice structures of silicon in Ga1−x Alx As alloys [48] for low-Al composition
(x) has provided indirect support for the reliability of the calculated phonons
of GaAs using RIM11.
Lattice Distortion
where ∆Eb1 and ∆Eb2 are the changes in the energy of the bonds caused by dis-
tortion in the nearest-neighbour and next nearest-neighbour atom positions,
respectively. The local distortion is then calculated by taking, ∂∆Eb /∂∆d = 0.
For GaAs:N, the calculated variations of impurity–host ∆Eb1 , host–host ∆Eb2 ,
and total ∆Eb change in bond energies as a function of distortion ∆d/do is
displayed in Fig. 9.11.
240 D.N. Talwar
0.4
0.2 ∆Eb2
Energy (eV)
−0.2
−0.4 ∆Eb = ∆Eb1 + ∆Eb2
−0.6
GaAs : N
−0.8
∆Eb 1
−1
−1.2
−0.25 −0.2 −0.15 −0.1 −0.05 0
∆ d / do
Fig. 9.11. The variation of change in impurity–host ∆Eb1 , host–host ∆Eb2 , and total
∆Eb bond energy vs. ∆d/do for GaAs:N
The BOM calculation of ∆d/do (= −0.18) for NAs in GaAs (see Fig. 9.11),
estimated at the minimum of total change in bond energy, is used to evaluate
the stiffening in the force constant variation parameter for the perturbation
matrix P of the Green’s function theory. Similar calculations were performed
for AlAs:N, InAs:N, GaAs:In, etc.
If the impurity space is confined to its nearest neighbours, then the simplest
case of an isolated N occupying the As site (NAs ) in GaAs (InAs or AlAs) with
Td symmetry (see Fig. 9.9a) would involve only five atoms. In the framework of
an RIM11, the two nearest-neighbour force constants A and B will be modified
to A (≡ A + ∆A) and B (≡ B + ∆B) or A and B , if the impurity M1I (or
M2I ) occupies the Ga (or the As) site in GaAs, respectively. The changes are
described by the parameters [48]
or
The displacements involving impurity atom and its four nearest neighbours
can be classified by the irreducible representations of the defect point group
Td , i.e. ΓTd ≡ A1 ⊕ E ⊕ F1 ⊕ 3F2 . Transformation to the symmetry-adapted
displacements can be used to reduce the defect and the Green’s functions
matrices to block diagonal form. We then solve (9.1) in the A1 , E, F1 , and F2
irreducible representations of the Td point group, as a function of the force
constant change parameter t or u for isolated defects occupying either the Ga
(In or Al) or the As site, respectively.
9 Assessing the Preferential Chemical Bonding of Nitrogen 241
650
11
B
600 12
C
Local mode frequency (cm−1) 550
13
C
Ga As
14
500 N
15
N
450 28
Si
30
400 Si
31
P
350 32
S
300
−1 −0.8 −0.6 −0.4 −0.2 0 0.2 0.4 0.6
Force constant change parameter, u
Table 9.3. Comparison of the calculated localized vibrational modes due to closest
mass N (isoelectronic) and C (acceptor) (Be (acceptor), B (isoelectronic) defects)
occupying As, P, and Sb site (Ga, In) in various III–V compound semiconductors
Several interesting trends are noted in the force constant variations for
closest mass CV (a− ) and NV (i) occupying the V site; and BeIII (a− ) and BIII (i)
occupying the III site in III–V compounds:
(a) For CAs acceptors in III–As, we find an increase in the force constant
of around 35% with respect to the closest mass isoelectronic NAs while
for BeIII and MgIII acceptors, there is a decrease in the values of force
constants by ∼30% from those of the closest mass isoelectronic BIII and
AlIII impurities, respectively.
(b) For SiIII donors in III–V compounds, on the other hand, we have found
an increase in the force constant (∼20%) with respect to the closest mass
isoelectronic AlIII .
It is worth mentioning that the above trends in the force constant variations
are found independent of the long-range Coulomb interactions. We strongly
believe, however, that the charged impurities in semiconductors affect only the
short-range forces via lattice relaxations and by the redistribution of the bond
electron charge density. The increase (decrease) in the bonding force constant
9 Assessing the Preferential Chemical Bonding of Nitrogen 243
Fig. 9.13. (a) Force constant variation correlation with bond covalency (ionicity) in
iso-row Ge–GaAs–ZnSe semiconductors and (b) charged impurities occupying the
cation and anion sites in GaAs
244 D.N. Talwar
Again, from the analysis of LVM for BAs in GaAs (see Table 9.3), our
Green’s function calculations provided force constant variation similar to CAs .
Since BAs is expected to act as a double acceptor (a2− ) while CAs is a single
acceptor (a− ), this result seems a little odd. A possible explanation for this
result might be that the extra charge on the BAs atom in GaAs is widely
distributed so that as far as the force constant with the nearest neighbours
is concerned, any adjustment by the redistribution of electron charge density
between the BAs and Ga bond is similar to CAs –Ga bond. Again, since the
gallium neighbours of the BAs and CAs atoms have different combinations of
69
Ga and 71 Ga isotopes, nine LVM frequencies are expected [52] from five dif-
ferent isotopic centres for each impurity configuration. High-resolution FTIR
spectroscopy [62], carried out at 4.2 K for BAs in Ga-rich GaAs after a small
dose of 2 MeV electron irradiation, has allowed five lines to be detected, of
which two are not fully resolved. Similar observation of Ga-isotopic fine struc-
tures of CAs LVMs provided strong corroboration for the site selectivity of C
in GaAs [39] and was successfully analysed by the Green’s function theory.
Finally, it is worth mentioning that the present Green’s function analysis of
LVMs for isolated substitutional defects in semiconductors is not applicable
for defects occupying the off-centre sites (e.g. oxygen in GaAs [63, 64]).
In Ga-rich (In-rich) Ga1−x Inx Ny As1−y layers with x ≤ 0.12 (x ≥ 0.9), one
can realize the possibility of a nearest-neighbour pair-defect 14 NAs –InGa (or
14
NAs –GaIn ) of trigonal symmetry (C3v ) (see Fig. 9.9b), in which one of
the four Ga (In) atoms in the neighbourhood of 14 NAs at site 2 (say) is
replaced by a larger size and heavier In, i.e. InGa (or a smaller size and lighter
GaIn ) atom at site 1. From group theoretical arguments, the vector spaces
formed by the displacement of “impurity centre” and its nearest neighbours
in the 14 NAs InGa (1)Ga3 (14 NAs GaIn (1)In3 ) configuration will be transformed
according to the irreducible representations: ΓC3v = 6A1 ⊕ 2A2 ⊕ 8E. As the
nearest-neighbour pair defect does not move in the A2 representation, only A1
and E types of modes are optically active. Clearly, defining the perturbation
matrix to study the vibrational modes of a pair defect of C3v symmetry is
more challenging than the Td case. In this configuration, one needs a mini-
mum of two (u and t) perturbation parameters. To make the P matrix more
meaningful, one may also consider a direct interaction Γ12 between the two
impurities in terms of the force variation F12 (≡ u + t − ut + Γ12 ).
For the 14 NAs –InGa pair defect in Ga-rich GaInNAs layers grown on (100)
GaAs, we have defined the appropriate parameters by including corrections
to the force constants of isolated defects after studying lattice relaxation of
14
NAs and InGa in GaAs, respectively. In the framework of a first principle
BOM, we estimated ∼5% increase in the N–In (N–Ga) bond lengths from that
of the pure InN 2.14 Å (GaN 1.95 Å) bond, in good agreement with the results
obtained by an empirical pseudo-potential method using large atomistically
9 Assessing the Preferential Chemical Bonding of Nitrogen 245
relaxed super-cells [65]. The effects of lattice relaxation lead to 2.7% softening
in the N–In bond, a negligible change in the In–As and 8.6% stiffening in the
N–Ga bond, respectively. Although the effects of mismatch strain in the GaIn-
NAs layers are difficult to incorporate, we have chosen, however, in our study
u = Γ12 to keep the number of force constant parameters a minimum. For
the 14 NAs –InGa pair, our results of Green’s function calculations have clearly
revealed the splitting of a triply degenerate 14 NAs ∼471 cm−1 (In-free) F2 band
into a non-degenerate ∼462 cm−1 (A1 ) and a doubly degenerate ∼490 cm−1
(E) local mode in good agreement with the FTIR [31] (457 and 487 cm−1 )
and Raman [35] scattering (462 and 492 cm−1 ) data (see Table 9.4).
It is to be noted that the doubly degenerate E-mode describes the vibration
in the plane orthogonal to the N–In bond and is mainly affected by the force
constant between the N and Ga atoms, whereas the non-degenerate A1 -mode
represents the vibration of N atom along the bond of the defect pair and is
primarily influenced by the force constant between the N and In bond. Since
the N–Ga bond stretching force constant is larger than the force constant
Table 9.4. Comparison of the calculated local vibrational modes with the exper-
imental data of NAs involving In and Ga atoms in different configurations of Ga-
and In-rich Ga1−x Inx Ny As1−y alloys
between the In and N bond, the frequency of the E-mode is found higher for
the In–N pair in the 14 NAs InGa (1)Ga3 configuration than the A1 -mode. The
lower value of the A1 -mode frequency is caused by the collective effects of
larger indium mass and weaker force constant between the N and In bonds.
Again by changing the N-isotopic mass and keeping the same force constant
variations in the perturbation matrix, our Green’s function calculations for
the 15 NAs InGa (1)Ga3 configuration predicted a non-degenerate A1 -mode near
448 cm−1 and a doubly degenerate E-mode near 476 cm−1 , respectively.
Results of similar calculations performed for the LVMs of NAs –GaIn pair
in In-rich (Ga-rich) Ga1−x Inx Ny As1−y alloys are also included in Table 9.4.
It is to be noted that a low-temperature Raman study by Wagner et al. [42]
has revealed only Ga–N and In–N like modes near 470 and 443 cm−1 , respec-
tively, on high-In content Ga1−x Inx Ny As1−y (with y ≤ 0.012 and x ≥ 0.92)
grown by rf-nitrogen plasma source MBE on InP substrate. More recent FTIR
study [66] at 77 K of the local modes in 14 N-implanted In0.26 Ga0.74 As layers
annealed at 700–900◦ C has observed a weaker band near 510 cm−1 (A1 ) in
addition to the two modes detected in the Raman spectroscopy near 472
and 447 cm−1 . Unlike 14 NAs –InGa pair modes in Ga-rich material, the solu-
tion of (9.1) for the LVMs of 14 NAs –GaIn pair in In-rich GaInNAs provided
lower value of InN-like doubly degenerate mode and a higher GaN-like non-
degenerate mode [37] in good agreement with the recent FTIR [66] and Raman
scattering [67] data.
In Ga- (In-) rich Ga1−x Inx Ny As1−y alloys with In (Ga) contents ranging
between 0.1 and 0.3, one may speculate the possibility of the formation
of a second nearest-neighbour complex NAs InGa (2)Ga2 (NAs GaIn (2)In2 ) of
orthorhombic symmetry (C2v ). For such a defect centre in GaAs (say), we
consider replacing two of the four Ga atoms (at sites 1 and 6) by two larger
size and heavier InGa atoms in the vicinity of NAs (on site 2) (or in InAs sub-
stituting two of the four In atoms by smaller size and lighter GaIn atoms; see
Fig. 9.9c). In this configuration, the vector spaces formed by the displacements
of impurity atoms and their nearest neighbours transform according to the
irreducible representations: ΓC2v = 10A1 ⊕6A2 ⊕8B1 ⊕9B2 , where A1 , B1 , and
B2 type of vibrations are optically active. For studying the vibrational modes
of such a defect centre (NAs InGa (2)Ga2 ), by the Green’s function theory, we
require at least two u, t (= v) force constant change parameters. To make the
perturbation matrix P meaningful, one may also consider direct interactions
between impurities at sites 1–2 (Γ12 ) and 2–6 (Γ26 ) via the additional force
variation parameters F12 (≡ u + t − ut + Γ12 ) and F26 (≡ u + v − uv + Γ26 ),
respectively.
From the sparse data of N-related modes in Ga-rich GaInNAs and rec-
ognizing the fact that the observed Raman bands at frequencies higher than
9 Assessing the Preferential Chemical Bonding of Nitrogen 247
500 cm−1 are associated with the second-order scattering of GaAs-like opti-
cal modes, we first choose only two (u and t = v) perturbation parameters
for the complex centre of C2v symmetry. In the framework of BOM, our
study of lattice relaxation for the NAs InGa (2)Ga2 centre provided 3.6% soft-
ening in the N–In bonds and a small stiffening 9.6% in the N–Ga bonds,
respectively. With appropriate adjustments of force constants (t = v and
u) in the P matrix, the Green’s function theory provided a complete lift-
ing of the degeneracy of the 14 NAs (15 NAs ) local mode – predicting three
distinct bands near 481 (B1 ), 457 (A1 ), and 429 (B2 ) cm−1 (468, 444, and
417 cm−1 ) for 14 NAs InGa (2)Ga2 (15 NAs InGa (2)Ga2 ), respectively. Theoretical
results (see Table 9.4) are found to be in good agreement with the Raman
scattering [42] data as well as with two FTIR [66] bands observed near 480
and 454 cm−1 . In the B1 - and B2 -modes as N–Ga or N–In bonds are involved,
we find their frequencies lower than the values in the C3v case for the present
choice of parameters. The reason for the lower values is believed to be the
combined effects caused by the involvement of a second heavier In mass in the
NAs InGa (2)Ga2 cluster and by the weakening of force constant between the N
and In bonds.
It is worth mentioning, however, that the frequencies of impurity modes for
this configuration are found strongly dependent on the choice of the force con-
stant change parameters especially the values between Ga and N bonds. For
instance by retaining the same set of force constant change parameter t (= v)
between In and N bonds, and increasing the force constant u (= Γ12 = Γ26 )
between the N and Ga bonds from 9.6 to 16%, our Green’s function theory
for the 14 NAs InGa (2)Ga2 (15 NAs InGa (2)Ga2 ) cluster has provided shifts to
the LVM frequencies to 499 (B1 ), 479 (A1 ), and 458 (B2 ) cm−1 (485 (B1 ),
465 (A1 ), and 445 (B2 ) cm−1 ) in accord with at least two of the observed
Raman [42] and FTIR bands near 480 and 454 cm−1 [66]. Although the calcu-
lated local mode at 499 cm−1 falls near the frequency range dominated by the
second-order GaAs-like optical phonons, a broad IR band recently observed
near 521 cm−1 after annealing at 900◦C [66] has been assigned as a B1 -mode.
If one accepts the feature near 521 cm−1 to be a true local mode and not
as an artefact, then in In-rich GaInNAs one should have also observed an
A1 -mode near 500 cm−1 for the nearest-neighbour 14 NAs –GaIn pair of C3v
symmetry.
Accepting such a scenario requires an additional 7% stiffening in the Ga–N
bond for defining the perturbation matrix in the Green’s function theory for
the 14 NAs GaIn (1)In3 (15 NAs GaIn (1)In3 ) nearest-neighbour pair in the InAs
lattice. This additional stiffening in the Ga–N bond shifts the LVMs of
14
NAs GaIn (1)In3 (15 NAs GaIn (1)In3 ) pair to 496 cm−1 (A1 ) and 445 cm−1 (E)
(482 cm−1 (A1 ) and 430 cm−1 (E): see Table 9.4), respectively. For the
14
NAs GaIn (2)In2 (15 NAs GaIn (2)In2 ) C2v cluster in In-rich materials, our study
required even larger stiffening (10%) in the Ga–N force constants for matching
the LVMs modes closer to those of the 14 NAs InGa (2)Ga2 (15 NAs InGa (2)Ga2 )
in Ga-rich GaInNAs alloys.
248 D.N. Talwar
N, three new Raman lines were detected earlier near 425, 458, and 480 cm−1 .
Although the origin of 425 cm−1 feature was not fully understood, the authors
of [42] speculated it, however, either to the involvement of a nitrogen dimer,
i.e. NN on the As site (NNAs ), or to a Ga–N mode in the N-rich domain. The
other two lines observed near 458 and 480 cm−1 were thought to be related
to the nearest-neighbour 14 NAs InGa (1)Ga3 pair of C3v symmetry [42]. More
recently, two modes in 14 N-implanted Ga0.84 In0.16 As layers observed near 454
and 480 cm−1 by infrared spectroscopy [66] after RTA treatment at 700 and
800◦ C are attributed to the 14 NAs InGa (2)Ga2 pair of C2v symmetry. Again
after annealing at 900◦ C, the intensities of lines near 454 and 480 cm−1 are
found to decrease with the emergence of a new broadband at 521 cm−1 .
From a theoretical stand point and using a simplified BOM, we have esti-
mated qualitatively the lattice relaxation effects caused by 14 NAs InGa (2)Ga2
in GaAs lattice. The study provided a small softening (3.6%) and stiffening
(9.6%) in the N–In and N–Ga bonds, respectively. By including appropriate
force variations in the perturbation matrix for the 14 NAs InGa (2)Ga2 cen-
tre, our Green’s function theory predicted three distinct local modes near
429 (B2 ), 457(A1 ), and 481 (B1 ) cm−1 in good agreement with the Raman
data [42]. We have also noted, however, that the mode frequencies of the C2v
centre are strongly dependent on the choice of force constants between the
N and Ga bonds. By increasing the N–Ga bond strength from 9.6 to 16%
and retaining the N–In value as before, our calculations provided shifts in the
LVMs (458 (B2 ), 479 (A1 ), and 499 (B1 ) cm−1 ). Although two of the three
modes (B2 and A1 ) are in conformity with the recent FTIR [66] and Raman
data [42], the highest calculated frequency of B1 -mode (499 cm−1 ) falls in the
spectral range dominated by the second-order optical phonons of GaAs. In N-
implanted InGaAs layers, we believe that the feature observed near 521 cm−1
after RTA at 900◦ C is too broad and higher in frequency to be associated
with the B1 -mode. However, if it is a true local mode and not an artefact,
as suggested [66], then the large width caused by the shortening of lifetime is
probably related to its strong interaction with the lattice phonons. To check
the validity of our theoretical conjectures and to test the experimental [66]
claims of the true local modes near or above ≥500 cm−1 , we strongly feel the
need of more sophisticated calculations by first principle methods as well as
the necessity of additional experiments by using Raman scattering and/or IR
spectroscopy.
Similar to Ga-rich Ga1−x Inx Ny As1−y alloys, we have also anticipated
the formation of nearest-neighbour pairs 14 NAs GaIn (1)In3 of C3v and
14
NAs GaIn (2)In2 complexes of C2v configurations in In-rich materials (1.0 >
x ≥ 0.7, y > 0.005). As compared to 14 NAs InGa (1)Ga3 cluster of C3v sym-
metry where E > A1 , the role of A1 - and E-modes is, however, reversed (i.e.
E < A1 ) in 14 NAs GaIn (1)In3 . The A1 -mode in 14 NAs GaIn (1)In3 – caused by
the vibration of nitrogen along the N–Ga axis – is influenced by a strong
N–Ga bond and three weak N–In bonds. The E-mode – influenced by the
vibration of nitrogen atom perpendicular to the axis – is dependent upon
250 D.N. Talwar
the weaker N–In bonds. Therefore, one suspects the frequency of E-mode to
fall near 443 cm−1 and the A1 -mode to lie at a value close to or higher than
≥470 cm−1 (see Table 9.4). Quite recently, low-temperature studies of local
modes have been reported in Ga1−x Inx Ny As1−y with y ≤ 0.012 and x ≥ 0.92
using Raman scattering [42] and by FTIR in N-implanted Inx Ga1−x As layers
(with x = 0.26, 0.53, 0.75, and 1) annealed at 700–900◦C [66]. Although the
two experimental results have provided different values of LVMs, both have
confirmed the preferential bonding with a significant fraction of substitutional
nitrogen attached at least to one Ga neighbour.
In summary, our comprehensive Green’s function analyses of the Raman
and FTIR data on thermally annealed Ga- (In-) rich GaInNAs samples have
provided a strong support for the creation of preferential N–In (N–Ga) bond-
ing beyond what is statistically expected from a random alloy. By using
simple perturbation models, we have predicted nine local vibrational modes
for five different N–Ga4−m Inm clusters with m = 0, 1, 2, 3, 4. In Raman
studies, although no evidence of N-centred clusters involving more than one
In (Ga) atoms is found, the possibility of N–In2 Ga2 clusters cannot be com-
pletely ruled out, however, in GaInNAs samples with higher In (Ga) contents
and/or upon RTA at 600–900◦ C [66]. In the MBE-grown low-In content
layers of Ga1−x Inx Ny As1−y with x = 0.058 and y = 0.028, the annealing-
induced blueshift of the band gap has been suggested to the changes in recent
photo-reflectance resonances [69] associated with three different N-centred
short-range order clusters (N–Ga4 , N–Ga3 In1 , and N–Ga2 In2 ). The mecha-
nism for the redistribution of N environment from Ga- (In-) ligand rich sites
to In- (Ga-) ligand rich sites is not well understood, the hopping of N or
vacancy-assisted migration of In and Ga atoms at elevated temperatures is
speculated for the rearrangement of group III–N–As bonding. To check our
theoretical conjectures, more systematic work on the impurity vibrational
modes is necessary by using sophisticated first principle methods as well as
spectroscopic techniques by using Raman and FTIR methods to elucidate the
local coordination of N and In (Ga) in Ga- (In-) rich dilute GaInNAs alloys
under different growth conditions with varied In (Ga) contents and annealing
procedures.
References
1. For recent reviews, please see: “Special issue: III–V–N Semiconductor Alloys”,
J.W. Ager, W. Walukiewicz (eds.), Semicond. Sci. Technol. 17 (2002); and “Spe-
cial issue: The Physics and Technology of Dilute Nitrides”, N. Balkan (ed.), J.
Phys. Condens. Matter 16 (2004)
9 Assessing the Preferential Chemical Bonding of Nitrogen 251
27. R.C. Newman, Adv. Phys. 18, 545 (1969); ibid., Semicond. Semimet. 38, 117
(1993); ibid., Semicond. Sci. Technol. 9, 1749 (1994)
28. M. Stavola, Semicond. Semimet. B 51, 153 (1999)
29. M.D. McCluskey, Appl. Phys. Rev. 87, 3593 (2000)
30. H.C. Alt, J. Phys. Condens. Matter 16, S3037 (2004); ibid., H.C. Alt, Y.V.
Gomeniuk, B. Wiedemann, Phys. Rev. B 69, 125214 (2004); ibid., Semicond.
Sci. Technol. 18, 303 (2003); ibid., Appl. Phys. Lett. 77, 3331 (2000); ibid., Mat.
Sci. Forum 258–263, 867 (1997)
31. S. Kurtz, J. Webb, L. Gedvilas, D. Friedman, J. Geisz, J. Olsen, R. King, D.
Joslin, N. Karam, Appl. Phys. Lett. 78, 748 (2001)
32. T. Prokofyeva, T. Sauncy, M. Seon, M. Holtz, Y. Qiu, S. Nikishin, H. Temkin,
Appl. Phys. Lett. 73, 1409 (1998)
33. M.J. Seong, M.C. Hanna, A. Mascarenhas, Appl. Phys. Lett. 79, 3974 (2001);
ibid., Semicond. Sci. Technol. 17, 823 (2002)
34. M.S. Ramsteiner, D.S. Jiang, J.S. Harris, K.H. Ploog, Appl. Phys. Lett. 84,
1859 (2004)
35. K. Köhler, J. Wagner, P. Gesner, D. Serries, T. Geppert, M. Maier, L. Kirste,
IEE Proc. Optoelectron. 151, 247 (2004); ibid., J. Appl. Phys. 90, 2576 (2004);
ibid., Solid State Electron. 47, 461 (2003); ibid., Mater. Res. Symp. Proc. 744,
627 (2003); ibid., Appl. Phys. Lett. 83, 2799 (2003); ibid., Appl. Phys. Lett. 80,
2081 (2002); ibid., J. Appl, Phys. 90, 5027 (2001); ibid., Appl. Phys. Lett. 77,
3592 (2000)
36. D. Strauch, B. Dorner, J. Phys. Condens. Matter 2, 1457 (1990)
37. D.N. Talwar, J. Appl. Phys. 99, 123505 (2006); ibid., IEE Proc. Circuits Devices
Syst. 150, 529 (2003)
38. A.A. Kachare, W.G. Spitzer, W.G. Kahan, F.K. Euler, T.A. Whatley, J. Appl.
Phys. 44, 4393 (1973)
39. W.M. Theis, K.K. Bajaj, C.W. Litton, W.G. Spitzer, Appl. Phys. Lett. 41, 70
(1982)
40. G.A. Gledhill, S.B. Upadhyay, M.J.L. Sangster, R.C. Newman, J. Mol. Struct.
247, 313 (1991)
41. R.J. Elliot, W. Hayes, G.D. Jones, H.F. MacDonald, C.T. Sennett, Proc. R.
Soc. A 289, 1 (1965)
42. J. Wagner, K. Köhler, P. Ganser, M. Maier, Appl. Phys. Lett. 87, 051913 (2005)
43. T. Kitatani, M. Kondow, M. Kudo, Jpn. J. Appl. Phys. 40, L750 (2001)
44. H.C. Alt, Y.V. Gomeniuk, Phys. Rev. B 70, 161314 (2004)
45. S. Baroni, S. De Girnocoli, A. Dal Corso, P. Gianozzi, Rev. Mod. Phys. 73, 515
(2001)
46. R. Jones, P.R. Briddon, in Identification of Defects in Semiconductors, ed. by M.
Stavola. Semiconductors and Semimetals, vol. 51A, Chap. 6 (Academic, Boston,
1998)
47. A.A. Maradudin, E.W. Montroll, C.H. Weiss, I.P. Ipatova, in Solid State Physics,
ed. by F. Seitz, D. Turnbull, H. Ehrenreich (Academic, New York, 1971)
48. D.N. Talwar, M. Vandevyver, Phys. Rev. B 40, 9779 (1989)
49. D.N. Talwar, Phys. Stat. Solidi C 4, 674 (2006)
50. W.A. Harrison, Electronic Structure and the Properties of Solids – The Physics
of Chemical Bond (Dover, New York, 1980)
51. K. Kunc, M. Balkanski, N. Nusimovici, Phys. Stat. Solidi B 72, 229 (1975)
52. D.N. Talwar, M. Vandevyver, W. Theis, K.K. Bajaj, Phys. Rev. B 33, 8525
(1986)
9 Assessing the Preferential Chemical Bonding of Nitrogen 253
53. B.D. Rajput, D.A. Browne, Phys. Rev. B 53, 905 (1996)
54. S.F. Ren, H. Chu, Y.C. Chang, Phys. Rev. B 37, 8899 (1988)
55. Z.V. Popovic, M. Cardona, E. Richter, D. Strauch, L. Tapfer, K. Ploog, Phys.
Rev. B 41, 5904 (1990); ibid., Phys. Rev. B 40, 1207 (1989)
56. D.N. Talwar, K.S. Suh, C.S. Ting, Philos. Mag. B 54, 93 (1986)
57. L. Bellomonte, J. Phys. Chem. Solids 38, 59 (1977)
58. M.D. McCluskey, E.E. Haller, P. Becla, Phys. Rev. B 65, 045201 (2001)
59. S. Najmi, X. Zhang, X.K. Chen, M.L.W. Thewalt, S.P. Watkins, Appl. Phys.
Lett. 88, 041908 (2006)
60. J.R. Chelikowsky, M.L. Cohen, Phys. Rev. B 14, 556 (1976)
61. R. Jones, S. Öberg, Phys. Rev. B 44, 3407 (1989); ibid., Semicond. Sci. Technol.
7, 855 (1992)
62. G.A. Gledhill, R.C. Newman, J. Woodhead, J. Phys. C: Solid State Phys. 17,
L301 (1984)
63. J. Schneider, B. Dischler, H. Seelewind, P.M. Mooney, J. Lagowski, M. Matsui,
D.R. Beard, R.C. Newman, Appl. Phys. Lett. 54, 1442 (1989)
64. H.C. Alt, Semicond. Sci. Technol. 6, B121 (1991)
65. P.R. Kent, A. Zunger, Phys. Rev. B 64, 115208 (2001)
66. H.C. Alt, Y.V. Gomenuik, G. Mussler, Semicond. Sci. Technol. 21, 1425 (2006)
67. M. Ramsteiner, G. Mussler, P. Kleinert, K.H. Ploog, Appl. Phys. Lett. 87,
111907 (2005)
68. A. Hashimoto, T. Kitano, K. Takahashi, H. Kawanishi, A. Patane, C.T. Foxon,
A. Yamamoto, Phys. State Solidi B 228, 283 (2001); ibid., Phys. State Solidi B
234, 915 (2002)
69. R. Kudrawiecz, J. Misiewicz, E.M. Pavelescu, J. Konttinen, M. Pessa, Acta
Phys. Pol. A 106, 249 (2004)